How our games use AI
Some of our games let you play in your own words. To understand what you type, your input is sent through our server to an external AI service. The AI only interprets your input — it never writes dialogue or anything else you see. This page lists which services we use, what is sent to them, and what each provider does with it.
Last updated: 26 October 2026
Choose your region
We use different providers depending on where the game runs. Pick the region that applies to you.
You are reading the global version. In China our games run entirely on Alibaba Cloud instead — both our server and the AI model are in China.
You are reading the China version. Outside China our games use a server on AWS and TypeSafe AI's Jev instead.
What the AI does in this game
In PLEASE EXPLAIN you interrogate suspects by typing your own questions instead of picking from preset options. The AI is used only to interpret what you meant. It does not generate dialogue or any other content you see — everything on screen is written in advance by our team.
- Reads the question or accusation you typed and decides which kind of move it is — for example, questioning an alibi or pointing out a contradiction with the evidence.
- Picks that answer from a fixed list of intent types designed in advance by our team.
- Write dialogue or story. Every character's response is written in advance by our team.
- Decide what happens. The game's rules decide which response you get.
- Generate images, voice or any other content.
Services we use
Our relay server runs on AWS. The AI model is TypeSafe AI's Jev — a decision model that picks an answer from options we give it and never writes text.
Everything runs inside China: our relay server on Alibaba Cloud, and the decision model on Alibaba Cloud Model Studio (Bailian). Like Jev, it only picks an answer from options we give it and never writes text. The version of the game distributed in China is a separate build that connects only to these services.
- What it handles
- Picks which of our intent types matches what you typed. It returns a decision, never text.
- Processed by
- TypeSafe AI, Inc., in the United States.
- What it handles
- Picks which of our intent types matches what you typed. It returns a decision, never text.
- Processed by
- Alibaba Cloud, in China.
How the game connects
The game never talks to the AI service directly. Everything goes through our server, which keeps the provider's API keys out of the game.
- When you type a question, the game sends your text and the current scene to our server.
- Our server adds what the AI needs to judge it — such as the list of intent types to choose from — and sends the request to the AI service for your region: Jev globally, the Bailian decision model in China. The provider's API keys stay on our server.
- The AI service returns only its decision — which intent type it picked — to our server, which passes it back to the game.
- The game then follows its own rules to show dialogue and events written by our team.
What is sent
- The text you typed.
- The current scene in the game, such as which case and which character you are questioning.
- Your name, email address or other contact details — the game never asks for them.
Please don't type your own or anyone else's personal information into the game.
What each provider does with it
Each provider sets its own terms for the data it receives. The summaries below are ours; the linked documents are authoritative.
| Provider | Used to train models | Retention | Official documents |
|---|---|---|---|
| TypeSafe AISystem One APIJev (jev-latest) | No TypeSafe's privacy policy states that it will not train or fine-tune any AI or machine-learning models on prompts or other input. | Kept for as long as reasonably necessary to provide the service, per TypeSafe's privacy policy. The service is hosted in the United States. | TypeSafe AI Privacy Policy ↗TypeSafe AI Terms of Use ↗Jev documentation ↗ |
| Alibaba CloudModel Studio (Bailian) — decision modeldecision-model-preview | No Alibaba Cloud states that it strictly protects data privacy and will never use your data for model training. | As required by applicable laws and regulations, Model Studio stores data generated when models are called. It is handled under the Model Studio service agreement. Processing takes place in China. | Model Studio compliance & privacy notice (Chinese) ↗Model Studio service agreement (Chinese) ↗Decision model documentation (Chinese) ↗ |
Based on information published by each provider as of October 2026.
Guardrails on what the AI can do
- Both AI models can only return one of the intent types we provide. They cannot produce text.
- Input that doesn't match any type gets a general response written by our team.
- AI-generated text is never shown on screen. If an interpretation ever feels wrong, tell us on our Discord server or by email and we will look into it.
What we keep
- Input logs — the text you typed and the decision that was returned — stored anonymously and not linked to your Steam account. We use them to improve how accurately the game understands you.
- Anonymous gameplay figures, such as clear times and the number of interrogations, used to balance the difficulty. These are not linked to your Steam account either.
- Payment information. Purchases are handled by Steam, so we never receive your card details.
Questions
If anything on this page is unclear, or you have a question about how our games use AI, get in touch — we are happy to help.